Image analysis – Applications – Biomedical applications
Reexamination Certificate
2000-11-22
2004-06-01
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
Biomedical applications
Reexamination Certificate
active
06744911
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to image processing and more particularly segmenting and classifying a region or data type of interest based on selected characteristics.
Identification of contiguous regions of the same material is known as segmentation. The automatic classification or segmentation of medical images into anatomical regions, for example bone, kidney or lung, is a difficult problem that is currently stalling the development of many clinical applications. While several techniques have been proposed over the years, very few techniques have achieved the high degree of automation and quality required by routine medical practice. Instead, most clinical applications rely on a large amount of user interaction. Thus, the process of segmentation is time-consuming, error prone, and subjective.
The human skeletal system is mainly comprised of two types of bone cortical bone and trabecular bone. Cortical bone is quite dense and typically located at the perimeter of a bone region. Trabecular bone is spongy and is typically located in interior bone regions. As a result, bone regions have several distinct characteristics that are nearly universal in the patient CT scan population. CT densities rapidly rise at bone boundaries (cortical) and then drop modestly within a few millimeters, thereby creating a “ring” effect in scan data. The structure of interior bone (trabecular) has a distinct spongy pattern that is characterized by constant change and relatively high frequency. The CT bone values range from 1200 to greater than 2000 Hounsfeld units for cortical bone, and trabecular ranging from 1000 to 1300 Hounsfeld units.
Although bone, and in particular cortical bone, presents itself as a high density signal which is higher than most any other region of the human body, segmentation of bone is not effectively performed with simple thresholding, a known technique of segmenting based on minimum or maximum CT numbers. Often, a foreign object or substance, such as a contrast agent, is introduced into a patient that exhibit CT densities that are indistinguishable from bone. This is most difficult in the case of a contrast enhanced CT scan where a liquid agent is injected into the blood stream resulting in vascular regions with CT numbers equal to or higher than bone.
The specific problem of identifying and removing bone from contrast enhanced computed tomography (CT) data illustrates many segmentation issues. CT values in contrast enhanced regions cannot be easily distinguished from other high density structures such as bone. Medical segmentation research has not revealed an automatic bone segmentation technique capable of robust identification of bone, despite the relative ease in which a human operator can perform the task. What is needed is a segmentation technique for classifying bone and other anatomical regions with high density characteristics in CT scan data, and to further distinguish bone from other regions.
BRIEF SUMMARY OF THE INVENTION
A method for segmenting an input volume in accordance with at least one selected characteristic comprises the steps of casting a plurality of rays along a plurality of views of the input volume, identifying regions corresponding to the selected characteristic within each ray, and processing the identified regions to generate an output volume indicating regions having the selected characteristic.
REFERENCES:
patent: 5226113 (1993-07-01), Cline et al.
patent: 5793375 (1998-08-01), Tanaka
patent: 5832134 (1998-11-01), Avinash et al.
patent: 5865750 (1999-02-01), Hatfield et al.
patent: 5891030 (1999-04-01), Johnson et al.
patent: 5959631 (1999-09-01), Knittel
patent: 6211884 (2001-04-01), Knittel et al.
patent: 6343936 (2002-02-01), Kaufman et al.
patent: 6353677 (2002-03-01), Pfister et al.
Avila Ricardo Scott
Volpe Christopher Richard
General Electric Company
Johns Andrew W.
Nakhjavan Shervin
Patnode Patrick K.
Testa Jean K.
LandOfFree
Tomographic segmentation does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Tomographic segmentation, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Tomographic segmentation will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3327921